Fix static Interventions Avoided per-100 scaling

[uriahf]

Summary

• fix Interventions Avoided so the full count-form expression is scaled per 100
• add direct numerical regression coverage at thresholds 0.25, 0.5, and 0.75 with false negatives present
• verify equivalence to 100 * (NB - NB_all) * (1-t)/t
• verify Treat All remains zero and Treat None retains its population-dependent per-100 path
• preserve existing public Decision Curve APIs and rendering behavior
• add the established changelog note

Statistical fix

Before:

100 * TN/N - FN/N * (1-t)/t

After:

100 * (TN/N - FN/N * (1-t)/t)

This matches the current R static Interventions Avoided definition.

Scope

No canonical InterventionsAvoidedV2Spec, rtichoke_viz, browser adoption, DecisionCurveV2Spec, ReportSpec, Quarto, Precision-Recall, or API redesign changes.

The static and time-dependent performance builders share the same metric-transform helper, so this arithmetic correction also applies mechanically to the existing time-dependent IA column; no time-specific derivation or API was changed.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-08-25 07:40 UTC